Progress In Electromagnetics Research M, Vol. 68, 173 180, 2018 Enhanced Characteristic Basis Function Method for Solving the Monostatic Radar Cross Section of Conducting Targets Jinyu Zhu, Yufa Sun *, and Hongyu Fang Abstract In this paper, an enhanced characteristic basis function method () is proposed to calculate the monostatic radar cross section (RCS) of electrical large targets efficiently. The enhanced characteristic basis functions (ECBFs) are defined by combining improved primary-characteristic basis functions (IP-CBFs) with the first level improved secondary-characteristic basis functions (IS-CBFs) for each block. IS-CBFs are obtained by substituting IP-CBFs for PCBFs in Foldy-Lax multiple scattering equation in which mutual coupling effects among all blocks can be included systematically. As a result, a small number of incident plane waves (PWs) is sufficient when dealing with large scale targets. The numerical results demonstrate that the computational efficiency in this paper is much higher than that of the improved primary-characteristic basis function method (IP-CBFM) without losing any accuracy. 1. INTRODUCTION The method of moments (MoM) [1] is accurate in the analysis of arbitrary three-dimensional electromagnetic scattering. However, with the increase of the size of the targets under analysis, the computational time and memory requirements of the conventional MoM are very expensive. Some fast algorithms such as multilevel fast multipole algorithm (MLFMA) [2], adaptive integral method (AIM) [3] and adaptive cross approximation (ACA) algorithm [4] are proposed to relieve this problem. However, these iterative methods suffer from convergence problems of ill-conditional matrices for electrical large targets. Moreover, when multiple incident plane waves (PWs) scattering problems are considered, the iterative procedure must be repeated for each PW. The characteristic basis function method (CBFM) [5] effectively reduces the dimension of impedance by dividing the target into several blocks, and high-level basis functions are generated to represent the electromagnetic characteristics of these blocks. These basis functions are referred to as the characteristic basis functions (CBFs), and the CBFs lead to a reduced. Since the method only requires the solution of small size equations, the LU decomposition method [6] can be applied to solve the reduced directly instead of the conventional iterative solving method. In [7], the excitation independent CBFM is proposed by obtaining a set of completely orthogonal primary CBFs (PCBFs), and the set of completely orthogonal PCBFs is used to construct the reduced. This method is widely applied in many aspects, especially for monostatic RCS. Based on the excitation independent CBFM, improved primary-characteristic basis function method (IP-CBFM) [8] is proposed to reduce the number of PWs by combining secondary CBFs (SCBFs) and PCBFs to generate improved primary CBFs (IP-CBFs) of each block. However, IP-CBFM still needs more PWs when dealing with large scale targets, which lead to increasing the time of CBFs generation and reduced filling. In this paper, an enhanced characteristic basis function method () is proposed to improve the efficiency of numerical calculation for the monostatic radar cross section (RCS). In, Received 27 February 2018, Accepted 25 April 2018, Scheduled 21 May 2018 * Corresponding author: Yufa Sun (yfsun ahu@sina.com). The authors are with the Key Lab of Intelligent Computing & Signal Processing, Ministry of Education, Anhui University, Hefei 230601, China.
174 Zhu, Sun, and Fang enhanced characteristic basis functions (ECBFs) consist of IP-CBFs and the first level improved secondary CBFs (IS-CBFs), where IS-CBFs are constructed by solving Foldy-Lax multiple scattering equation [9, 10] in which mutual coupling effects among all blocks can be included systematically. As both IP-CBFs and IS-CBFs are considered, a set of more complete CBFs can be obtained. As a result, a small number of PWs is sufficient when using this set of more complete CBFs. In addition, adaptive cross approximation-singular value decomposition (ACA-SVD) [11] is applied to compress these PWs to reduce the time generation of CBFs in this paper. Numerical results show that is more efficient than IP-CBFM and without losing any accuracy. 2. FORMULATION 2.1. The IP-CBFM The target is divided into M blocks, and the CBFs are composed of PCBFs and SCBFs in IP- CBFM. PCBF is the self-interaction component inside a block. SCBF indicates the mutual interaction component between block i and j (i =1, 2,...,M, j =1, 2,...,M). Let N θ and N ϕ indicate the number of PWs in θ and ϕ, respectively. Considering two kinds of, the total number of PWs is N p =2N θ N ϕ, which are arranged in a E Np.PCBFJ P for each block can be defined as follows: Z e J P = E Np (1) where Z e is an N i e Ni e extended self-impedance, and E Np is an Ni e N p. SCBF (i j) for each block can be expressed as J P ij Z e JP ij = Ze ij JP jj (i j) (2) where Z e ij is an N i e (N j N overlaping ij ) impedance of mutual interaction between block i and j, andn overlaping ij is the portion of the overlap between the extended block i and original block j. AccordingtoEqs.(1)and(2),IP-CBFJ IP can be obtained as Z e J P + Z e J P ij = Z e J P ij=z e J IP = E Np Z e ijj P jj (3) j=1 j i j=1 where J IP indicates roughly the correct current distribution of block i, and therefore, the number of PWs can be reduced. It is assumed that all of the blocks contain the same number K of IP-CBFs after ACA-SVD. Then, the dimension of reduced is KM. In addition to the generation of IP-CBFs, this method is the same as CBFM in [7]. 2.2. The Formulation of the When dealing with large scale targets, the number of incident PWs required to obtained valid results in IP-CBFM is still large. To solve this problem, IS-CBFs are added to construct ECBFs in IP-CBFM. According to the Foldy-Lax multiple scattering equation, the traditional first level SCBFs J S can be calculated by replacing incident field with scattered field due to the PCBFs on the all blocks except from itself. Z e J S = Z ij J P jj (4) j=1(j i) IS-CBF is the improvement of the traditional SCBF which is calculated by replacing PCBFs with IP- CBFs in the Foldy-Lax multiple scattering equation. By solving Eq. (5), the first level IS-CBFs J IS can be obtained as Z e J IS = Z ij J IP jj (5) j=1(j i) j=1 j i
Progress In Electromagnetics Research M, Vol. 68, 2018 175 where J IP and J IS are the last retained CBFs, which are called as ECBFs J E. As both IP-CBFs and IS-CBFs are considered, ECBFs are a set of more complete CBFs which can indicate the correct current distribution more accurately, and therefore, the number of PWs can be further reduced. Let H θ and H ϕ indicate the numbers of PWs in θ and ϕ, respectively. The total number of PWs is H p =2H θ H ϕ (H p N p ), which are arranged in a E Hp (i =1, 2,..., M). Moreover, incidents PWs with redundant information are used to calculate CBFs in IP-CBFM, which usually leads to increasing the solving time of CBFs. To mitigate this problem, ACA-SVD is applied to remove the redundancy of PWs as follows in this paper. First, the ACA of E Hp is expressed as E Hp E U E T V (6) Next, the QR decompositions [12, 13] of E U and E V are expressed as E U = Q i1 R 1 (7) E V = Q i2 R 2 (8) Then, R 1 R T 2 is recompressed by a truncated SVD R 1 R T 2 U is i Vi T (9) Substituting Eqs. (7) (9) into Eq. (6) E Hp Q i1 U i S i (Q i2 V i ) T (10) where the product of Q i1 U i is the last retained excitations after a truncated ACA-SVD. Finally, PCBFs J P on the block i can be obtained as Z e J P Q i1 U i (11) Substituting Eq. (11) into Eq. (3), the IP-CBFs J IP can be obtained as Z e JIP =Q i1u i j=1 j i Z e ij JP jj (12) Then, substituting J IP into Eq. (5), J IS can be obtained. For the sake of simplicity, it is assumed that all of the blocks contain the same number K E of IP-CBFs after ACA-SVD. Thus, the number of J E is 2K E, and the dimension of reduced is 2K E M(2K E M KM). Moreover, in order to reduce the time of CBFs generation, ACA-SVD is also used to compress incident PWs in IP-CBFM. However, although compression of incident PWs is used in the two methods, is more efficient than IP-CBFM in CBFs generation, since it requires only a small number of incident PWs. 3. NUMERICAL RESULTS In this section, several numerical examples were studied to demonstrate the accuracy and efficiency of the proposed method. The ACA and SVD threshold are chosen to be 10-3. All the results are computed on the Intel R Core TM i7-3820 3.60 GHz, 64 GB RAM PC. The compiler uses Code Blocks. The relative is defined as follows: / Err = 100 RCS χ RCS 2 RCS 2 (13) n n where RCS χ are the results of IP-CBFM and, and RCS are the simulation results of the software. First, the monostatic RCS of a PEC cylinder with 4 m length and 0.25 m radius at a frequency of 400 MHz is calculated. The geometry is divided into 2112 triangular patches, and the number of unknowns is 5696, which is divided into 16 blocks, and each block is extended by 0.15λ in all directions. The incidence angle is set to θ =0 180, ϕ =0. Incident PWs of PEC Cylinder are shown in Figure 1,
176 Zhu, Sun, and Fang Figure 1. Incident PWs of PEC Cylinder. Table 1. The analysis conditions of two methods. Method (P θ,p ϕ ) (Δθ inc, Δϕ inc ) Number of PWS (9, 4) (22.5,90 ) 72 (9, 8) (22.5,45 ) 144 (9, 2) (22.5, 180 ) 36 and the analysis conditions of two methods are shown in Table 1. For convenience, P θ,p ϕ are used to denote the numbers of incident PWs in θ and ϕ, respectively. Δθ inc, Δϕ inc indicate angle intervals of incident PWs. and represent the IP-CBFM with different numbers of incident PWs. The results of monostatic RCS in HH and VV s are calculated by IP-CBFM, and the proposed and are shown in Figure 2 and Figure 3. 20 0-20 20 0-20 Scattering angle (θ=0 180, ϕ=0 ) Scattering angle (θ=0 180, ϕ=0 ) Figure 2. HH monostatic RCS of Cylinder. Figure 3. Cylinder. VV monostatic RCS of
Progress In Electromagnetics Research M, Vol. 68, 2018 177 It can be seen in Figure 2, Figure 3 and Table 1 that with 72 PWs cannot obtain valid results. The results of and are in good agreement with those of. The calculation time and relative of the two methods are presented in Table 2. Table 2. The calculation time and relative of two methods. Method CBFs generation Reduced filling Total time (s) Dimension of reduced in HH in VV - - - - 33.61 18.26 13.05 107.99 149.16 1215 5.96 4.21 6.78 59.12 88.68 886 5.17 2.95 It can be found easily from Table 2 that the proposed outperforms the IP-CBFM in computation time and accuracy. The dimension of reduced is reduced by 27.1%, and computational efficiency is increased by 40.5%. Next, the monostatic RCS of 25 discrete PEC cubes with the length 1.0 m and the spacing between two cubes 1.0 m at the frequency of 250 MHz are computed and compared. The target is divided into 19000 triangular patches, and the number of unknowns is 28500. The target is just divided into 25 blocks. The incidence angle is set to θ =0 180, ϕ =0. Table 3 shows the analysis conditions of two methods. The results of monostatic RCS in HH and VV s calculated by IP-CBFM, and are shown in Figure 4 and Figure 5. The calculation time and relative of two methods are presented in Table 4. Table 3. The analysis conditions of two methods. Method (P θ,p ϕ ) (Δθ inc, Δϕ inc ) Number of PWS (20, 6) (9.5,60 ) 240 (16, 9) (12,40 ) 288 (9, 2) (22.5, 180 ) 36 Table 4. The calculation time and relative of two methods. Method CBFs generation Reduced filling Total Dimension of reduced in HH in VV - - - - 26.55 40.75 422.27 19594.14 20242.21 2809 7.07 11.93 209.83 6439.19 6815.69 1606 2.34 2.89 It can be seen in Figure 4, Figure 5 and Table 3 that with 240 PWs cannot obtain valid results, and even the results of cannot be in good agreement with those of, especially for VV. However, the results of are in good agreement with those of.
178 Zhu, Sun, and Fang 40 20 60 40 20 0 0 Scattering angle (θ=0 180, ϕ=0 ) Scattering angle (θ=0 180, ϕ=0 ) Figure 4. HH monostatic RCS of Discrete Cubes. Figure 5. VV monostatic RCS of Discrete Cubes. It can be easily seen from Table 4 that the proposed performs better than IP-CBFM in computation time and accuracy. Moreover, the dimension of reduced is reduced by 42.8%, and computational efficiency is increased by 66.3%. To further prove the accuracy and efficiency of the proposed method, the monostatic RCS of a PEC almond with 252.3744 mm at 6 GHz has been calculated. The target is divided into 20 blocks, and each block is extended by 0.2λ in all directions. The target is divided into 6236 triangular patches, and the total number of unknowns is 23161. The incidence angle is set to θ =90, ϕ =0 180. The analysis conditions of two methods are shown in the Table 5. Table 5. The analysis conditions of two methods. Method (P θ,p ϕ ) (Δθ inc, Δϕ inc ) Number of PWS (8, 6) (25.5,60 ) 96 (9, 7) (22.5,51.5 ) 126 (6, 4) (36,90 ) 48 The results of monostatic RCS in HH and VV s calculated by IP-CBFM, proposed and are shown in Figure 6 and Figure 7. The calculation time and relative of two methods are presented in Table 6. Table 6. The calculation time and relative of two methods. Method CBFs generation Reduced filling Total Dimension of reduced in HH in VV 150.98 853.89 1192.42 1379 4.92 9.27 191.65 1143.89 1540.11 1654 1.50 1.61 126.15 945.95 1259.83 1422 0.98 1.55
Progress In Electromagnetics Research M, Vol. 68, 2018 179-20 -40-20 -40-60 Scattering angle (θ=90, ϕ=0 180 ) -60 Scattering angle (θ=90, ϕ=0 180 ) Figure 6. HH monostatic RCS of Almond. Figure 7. Almond. VV monostatic RCS of It can be seen in Figure 6, Figure 7 and Table 5 that with 96 PWs cannot obtain valid results, especially for VV. However, the results of and are in good agreement with those of. It can be easily seen from Table 6 that the proposed performs better than IP-CBFM in computation time and accuracy, especially for HH. 4. CONCLUSION In this paper, an is proposed to solve the multiple incident PWs monostatic RCS of electrical large targets rapidly. ECBFs are defined and consist of IP-CBFs and first level IS-CBFs according to Foldy-Lax multiple scattering equation. The proposed method can reduce the time of CBFs generation and reduced filling significantly only with a small number of the incident PWs compared to IP-CBFM without losing any accuracy. The numerical results demonstrate that the proposed method is accurate and efficient. ACKNOWLEDGMENT This work is supported by the National Natural Science Foundation of China under Grant 61172020. REFERENCES 1. Harrington, R. F., Field Computation by Moment Methods, IEEE Press, New York, 1993. 2. Song, J., C. C. Lu, and W. C. Chew, Multilevel fast multipole algorithm for electromagnetic scattering by large complex objects, IEEE Trans. Antennas Propag., Vol. 45, No. 10, 1488 1493, 1997. 3. Bleszynski, E., M. Bleszynski, and T. Jaroszewicz, Adaptive integral method for solving largescale electromagnetic scattering and radiation problems, Radio Science, Vol. 31, No. 5, 1225 1251, 1996. 4. Zhao, K., M. N. Vouvakis, and J. F. Lee, The adaptive cross approximation algorithm for accelerated method of moments computations of EMC problems, IEEE Transactions on Electromagnetic Compatibility, Vol. 47, No. 4, 763 773, 2005. 5. Kwon, S. J., K. Du, and R. Mittra, Characteristic basis function method: A numerically efficient technique for analyzing microwave and RF circuits, Microwave & Optical Technology Letters, Vol. 38, No. 6, 444 448, 2003.
180 Zhu, Sun, and Fang 6. Pan, P. Q., A projective simplex algorithm using LU decomposition, Computers & Mathematics with Applications, Vol. 39, No. 1, 187 208, 2000. 7. Lucente, E., A. Monorchio, and R. Mittra, An iteration-free MoM approach based on excitation independent characteristic basis functions for solving large multiscale electromagnetic scattering problems, IEEE Trans. Antennas Propag., Vol. 56, No. 4, 999 1007, 2008. 8. Tanaka, T., Y. Inasawa, Y. Nishioka, and H. Miyashita, Improved primary-characteristic basis function method for monostatic radar cross section analysis of specific coordinate plane, IEICE Transactions on Electronics, Vol. E99, No. 1, 28 35, 2016. 9. Tsang, L., C. E. Mandt, K. H. Ding, and V. F. T. Article, Monte Carlo simulations of the extinction rate of dense media with randomly distributed dielectric spheres based on solution of Maxwell s equations, Optics Letters, Vol. 17, No. 5, 314 316, 1992. 10. Sun, Y. F., C. H. Chan, and R. Mittra, Characteristic basis function method for solving large problems arising in dense medium scattering, IEEE Antennas and Propagation Society International Symposium, Vol. 2, 1068 1071, 2003. 11. Bebendorf, M. and S. Kunis, Recompression techniques for adaptive cross approximation, Journal of Integral Equations & Applications, Vol. 21, No. 2009, 331 357, 2007. 12. Seo, S. M. and J. F. Lee, A single-level low rank IE-QR algorithm for PEC scattering problems using EFIE formulation, IEEE Trans. Antennas Propag., Vol. 52, No. 8, 2141 2146, 2004. 13. Burkholder, R. J. and J. F. Lee, Fast dual-mgs block-factorization algorithm for dense MoM matrices, IEEE Trans. Antennas Propag., Vol. 52, No. 7, 1693 1699, 2004.